2^0.7314n for exact Euclidean SVP, if you enjoy superlattices, discrete Gaussians, and still-astronomical space.

Committee picks stay hidden, yet censorship resistance survives, via an MPC inclusion-list protocol with deniability.

Power analysis on constant-time MiMC over BN254, plus RNR vs ISW masking results that finally care about overhead.

A practical NTT implementers' note on butterflies, twisting, incomplete transforms, and Good's trick, with the bounds spelled out.

Paper reformulates the wedge attack algebraically and extends it to multihomogeneous systems, then uses it to poke SNOVA.

Yet another way to squeeze more SIMD into BKS HSS: tensor encodings, authenticated automorphisms, and a near-maximal pack.

Lattice-based silent threshold crypto for DNF policies and TFHE, because apparently setup can be made annoying in bulk.

Shrinks reconstruction factors and modulus growth in threshold FHE over cyclotomic rings, with implementations that scale.

Million-scale private joins, apparently, now come with oblivious encrypted weight transfer and simulation security.

Garbled-circuit labels get a Bitcoin commitment that actually opens cleanly, with permissionless recovery.

How hard are Vandermonde knapsacks? This preprint says: softer than hoped, with attacks on PV and some vSIS commitments.

End-to-end ZK PDF417 decoding proves age/residency from C2PA images, because apparently barcodes needed a SNARK.

ZK for isogeny diamonds, four variants for different assumptions and degree-smoothness, because plain isogeny proofs were too normal.

Signature-free PQ IKE from an authenticated forward-secure KEM, with ROM/QROM proofs and state-exposure resistance.

LLM inference gets zkSNARKed at operator granularity, because apparently GPUs weren't opaque enough.

Linear-CRS batch arguments for NP in prime-order groups, courtesy of a composite-to-prime bilinear translation.

Paper on amortizing prover work across many independently checked proofs, via multi-verifier Fiat-Shamir and local folding.

Hawk key-recovery gets a smaller SVP: trace-zero cuts exact dimension from m/2+1 to 3m/8+1.

Arbitrary points, no sumcheck: DeepBrake folds RS proximity and evaluation binding into fewer queries and smaller proofs.

Small secrets turn sparse LWE/LPN into a Kikuchi-graph walk game with sample-runtime tradeoffs.

Passive key recovery for MQOM v2 lineage, reusing master seeds and shared roots to turn signatures into linear equations.

ePrint on a Hamming-weight correlation metric for key-dependent S-boxes, since affine-invariant scores still miss CPA leakage.

What if multi-signature privacy has to survive the public transcript too? This paper formalizes that and proves MuSig2-H.

Exact FHE inference gets runtime-configurable mixed precision, so bit-width isn’t baked into the keys anymore.

Can Freivalds plus code proximity tests make matrix multiplication checks SNARK-friendly? LAMP says yes, with O(tk) relation size.

Can Poseidon’s full rounds still fail to kill subspace trails? This paper says yes, with explicit MDS constructions.

Eprint on UC-secure threshold BLS key backup and recovery, with proactive refresh, because key custody was apparently too easy.

New conditional linear cryptanalysis framework with CLAT, then MILP searches for better trails on AES-like ciphers, because handwaving go...

Paper proving threshold Regev PKE under MLWE with noninteractive, CCA-secure decryption, because folklore was getting expensive.

Malicious NISC finally gets Yao-level communication, from random-bit OT correlations and a random oracle, no extra round trips.